[Diagnosismed-list] possible bug or feature?
Davis, Wade
davisjwa at health.missouri.edu
Thu Apr 22 17:38:09 CEST 2010
Thank you for making available your nice R package!
For one of my predictors, I encounter an error when the AUC is exactly 1. (It is a small sample of only 14). As you can see in my code below I get an error for the marker " IL1b" at time=5. I understand that no CI would be calculated in this, but the plot is not generated either.
I don't have any problems with the other marker/time combinations.
Also, have you considered adding an option for Pepe and Cai's empirical likelihood based CI? The attached paper shows it performs the best in most cases, and only bootstrap-based methods do better when AUC>0.95. Just something to think about...
Thanks for your comments and advice on this.
Wade
> aaron<-read.csv(file="F:\\Aaron Ericsson\\aaron.csv")
> print(aaron)
status y marker time truth
1 n 4966.5924 Il1b 3 -1
2 n 9883.7209 Il1b 3 -1
3 n 31528.6624 Il1b 3 -1
4 y 25501.6181 Il1b 3 1
5 y 142161.0169 Il1b 3 1
6 y 56346.1538 Il1b 3 1
7 y 671215.8809 Il1b 3 1
8 y 192592.5926 Il1b 3 1
9 y 385267.8571 Il1b 3 1
10 y 269010.0430 Il1b 3 1
11 y 67777.7778 Il1b 3 1
12 y 201988.6364 Il1b 3 1
13 y 91373.8019 Il1b 3 1
14 y 101904.7619 Il1b 3 1
15 n 487.7506 MIP1a 3 -1
16 n 1399.2248 MIP1a 3 -1
17 n 3590.2335 MIP1a 3 -1
18 y 2800.9709 MIP1a 3 1
19 y 26758.4746 MIP1a 3 1
20 y 23038.4615 MIP1a 3 1
21 y 164640.1985 MIP1a 3 1
22 y 29861.1111 MIP1a 3 1
23 y 46473.2143 MIP1a 3 1
24 y 72022.9555 MIP1a 3 1
25 y 3881.4815 MIP1a 3 1
26 y 28503.7879 MIP1a 3 1
27 y 15543.1310 MIP1a 3 1
28 y 11714.2857 MIP1a 3 1
29 n 77.2829 RANTES 3 -1
30 n 0.0000 RANTES 3 -1
31 n 1036.0934 RANTES 3 -1
32 y 851.1327 RANTES 3 1
33 y 3190.6780 RANTES 3 1
34 y 3984.6154 RANTES 3 1
35 y 35980.1489 RANTES 3 1
36 y 10185.1852 RANTES 3 1
37 y 8705.3571 RANTES 3 1
38 y 3902.4390 RANTES 3 1
39 y 4074.0741 RANTES 3 1
40 y 5473.4848 RANTES 3 1
41 y 3000.0000 RANTES 3 1
42 y 4685.7143 RANTES 3 1
43 n 11715.2104 Il1b 1 -1
44 n 60046.1894 Il1b 1 -1
45 n 77088.0361 Il1b 1 -1
46 y 66242.0382 Il1b 1 1
47 y 226551.2266 Il1b 1 1
48 y 1205766.7100 Il1b 1 1
49 y 519553.0726 Il1b 1 1
50 y 90790.8992 Il1b 1 1
51 y 635601.1184 Il1b 1 1
52 y 692076.2287 Il1b 1 1
53 y 122442.2442 Il1b 1 1
54 y 561085.9729 Il1b 1 1
55 y 501930.5019 Il1b 1 1
56 y 848101.2658 Il1b 1 1
57 n 1268.6084 MIP1a 1 -1
58 n 16859.1224 MIP1a 1 -1
59 n 60835.2144 MIP1a 1 -1
60 y 5171.9745 MIP1a 1 1
61 y 734487.7345 MIP1a 1 1
62 y 238532.1101 MIP1a 1 1
63 y 60055.8659 MIP1a 1 1
64 y 3661.9718 MIP1a 1 1
65 y 287045.6664 MIP1a 1 1
66 y 308926.7803 MIP1a 1 1
67 y 11782.1782 MIP1a 1 1
68 y 76923.0769 MIP1a 1 1
69 y 214414.4144 MIP1a 1 1
70 y 379746.8354 MIP1a 1 1
71 n 288.9968 RANTES 1 -1
72 n 923.7875 RANTES 1 -1
73 n 22404.0632 RANTES 1 -1
74 y 1968.1529 RANTES 1 1
75 y 53968.2540 RANTES 1 1
76 y 26430.7558 RANTES 1 1
77 y 5586.5922 RANTES 1 1
78 y 369.4475 RANTES 1 1
79 y 18434.2964 RANTES 1 1
80 y 40621.8656 RANTES 1 1
81 y 1790.4290 RANTES 1 1
82 y 3400.4525 RANTES 1 1
83 y 22535.3925 RANTES 1 1
84 y 15875.5274 RANTES 1 1
85 n 27902.3747 Il1b 5 -1
86 n 10582.0106 Il1b 5 -1
87 n 38611.1111 Il1b 5 -1
88 y 234274.1935 Il1b 5 1
89 y 583682.0084 Il1b 5 1
90 y 42348.7544 Il1b 5 1
91 y 438288.2883 Il1b 5 1
92 y 191954.0230 Il1b 5 1
93 y 456976.7442 Il1b 5 1
94 y 634146.3415 Il1b 5 1
95 y 79090.9091 Il1b 5 1
96 y 160169.4915 Il1b 5 1
97 y 137213.7405 Il1b 5 1
98 y 223870.9677 Il1b 5 1
99 n 1269.9501 MIP1a 5 -1
100 n 1023.8663 MIP1a 5 -1
101 n 3119.2661 MIP1a 5 -1
102 y 19849.4826 MIP1a 5 1
103 y 22765.2174 MIP1a 5 1
104 y 4586.5922 MIP1a 5 1
105 y 34254.4732 MIP1a 5 1
106 y 17162.9779 MIP1a 5 1
107 y 27427.6860 MIP1a 5 1
108 y 72821.7426 MIP1a 5 1
109 y 3624.4541 MIP1a 5 1
110 y 7972.9730 MIP1a 5 1
111 y 7472.3247 MIP1a 5 1
112 y 1603.3654 MIP1a 5 1
113 n 140.8978 RANTES 5 -1
114 n 0.0000 RANTES 5 -1
115 n 324.7706 RANTES 5 -1
116 y 680.1505 RANTES 5 1
117 y 2208.6957 RANTES 5 1
118 y 0.0000 RANTES 5 1
119 y 2775.3479 RANTES 5 1
120 y 547.2837 RANTES 5 1
121 y 1173.5537 RANTES 5 1
122 y 2334.1327 RANTES 5 1
123 y 1213.9738 RANTES 5 1
124 y 417.2973 RANTES 5 1
125 y 492.6199 RANTES 5 1
126 y 0.0000 RANTES 5 1
>
> aaron.df<-data.frame(aaron, "truth"=as.numeric(aaron$status)-1)
>
>
> library(DiagnosisMed)
>
>
> ROC(test=aaron.df[aaron.df$time==5&aaron.df$marker=="Il1b","y"],
+ gold=aaron.df[aaron.df$time==5&aaron.df$marker=="Il1b","truth.1"],
+ CL = 0.95)
Error in `row.names<-.data.frame`(`*tmp*`, value = c("Max. Accuracy", :
invalid 'row.names' length
>
J. Wade Davis, PhD
Assistant Professor
187 Galena DC 018.0
University of Missouri
Columbia, MO 65212
Phone: (573) 882-0770
Fax: (573) 884-4196
MU Biostatistics Group<http://som.missouri.edu/Research/BioStat/Default.aspx>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.r-forge.r-project.org/pipermail/diagnosismed-list/attachments/20100422/632b672c/attachment-0001.htm
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Comparison of non-parametric confidence intervals for the area under the ROC curve of a continuous-scale diagnostic testQIN.pdf
Type: application/pdf
Size: 170732 bytes
Desc: Comparison of non-parametric confidence intervals for
the area under the ROC curve of a continuous-scale diagnostic testQIN.pdf
Url : http://lists.r-forge.r-project.org/pipermail/diagnosismed-list/attachments/20100422/632b672c/attachment-0001.pdf
More information about the Diagnosismed-list
mailing list